The purpose of this study was to use high-resolution magnetic resonance (MR) imaging combined with image analysis to investigate the three-dimensional (3D) trabecular structure, anisotropy, and connectivity of human vertebral, femoral, and calcaneal specimens. The goal was to determine whether:​ (a) MR-derived measures depict known skeletal-site-specific differences in architecture and orientation of trabeculae;​ (b) 3D architectural parameters combined with bone mineral density (BMD) improve the prediction of the elastic modulus using a fabric tensor formulation;​ (c) MR-derived 3D architectural parameters combined with BMD improve the prediction of strength using a multiple regression model, and whether these results corresponded to the results obtained using higher resolution depictions of trabecular architecture. A total of 94 specimens (12 ⨉ 12 ⨉ 12 mm cubes) consisting of trabecular bone only were obtained, of which there were 7 from the calcaneus, 15 from distal femur, 47 from the proximal femur, and 25 from the vertebral bodies. MR images were obtained using a 1.5 Tesla MR scanner at a spatial resolution of 117 ⨉ 117 ⨉ 300 μm. Additionally, BMD was determined using quantitative computed tomography (QCT), and the specimens were nondestructively tested and the elastic modulus (YM) was measured along three orthogonal axes corresponding to the anatomic superior-inferior (axial), medial-lateral (sagittal), and anterior-posterior (coronal) directions. A subset of the specimens (n = 67) was then destructively tested in the superior-inferior (axial) direction to measure the ultimate compressive strength. The MR images were segmented into bone and marrow phases and then analyzed in 3D. Ellipsoids were fitted to the mean intercept lengths, using single value decomposition and the primary orientation of the trabeculae and used to calculate the anisotropy of trabecular architecture. Stereological measures were derived using a previously developed model and measures such as mean trabecular width, spacing, and number were derived. Because the spatial resolution of MR images is comparable to trabecular bone dimensions, these measures may be subject to partial volume effects and were thus treated as apparent measures, such as BV/TV, Tb.Sp, Tb.N, and Tb.Th rather than absolute measures, as would be derived from histomorphometry. In addition, in a subset of specimens, the Euler number per unit volume was determined to characterize the connectivity of the trabecular network. There were significant differences in the BMD, trabecular architectural measures, elastic modulus, and strength at the different skeletal sites. The primary orientation axes for most of the specimens was the anatomic superior-inferior (axial) direction. Using the fabric tensor formulation, in addition to BMD, improved the prediction of YM (SI), while including some of the architectural parameters significantly improved the prediction of strength. In comparing MR-derived 3D measures with those obtained from 20 μm optical images (n = 18;​ 9 vertebrae, 9 femur specimens), good correlations were found for the apparent Tb.Sp and Tb.N, moderate correlation was seen for the apparent BV/TV, and poor correlation was found for the apparent Tb.Th. Using these higher resolution images, the fabric tensor formulation for predicting the elastic modulus also showed improved correlation between the measured and calculated modulus in the axial (SI) direction. In summary, high-resolution MR images may be used to assess 3D architecture of trabecular bone, and the inclusion of some of the 3D architectural measures provides an improved assessment of biomechanical properties. Further studies are clearly warranted to establish the role of architecture in predicting overall bone quality, and the role of trabecular architecture measures in clinical practice. Currently, MR techniques may not be used instead of histomorphometry;​ however, they may provide an ideal platform for assessing trabecular architecture in vivo, at multiple skeletal sites longitudinally, and assist in understanding the etiology of osteoporotic and aging changes, for studying osteoporosis progression and therapeutic efficacy.